DEEP LEARNING TECHNIQUES FOR THE CATEGORY OF INDIAN CURRENCY

Abstract

The use of machines has replaced human labor in almost every industry due to technological advancements and development. Several areas of banking and automated sales systems may benefit from currency paper recognition. Accurately identifying paper currency notes is a must in today's age of automated current repeating systems. Unfortunately, as currency notes get worn and damaged, machines typically struggle to detect and recognize them in circulation. Predicting and analyzing real money notes is challenging for visually impaired persons without any technical help or aid. With the use of these models, the identification accuracy of currency notes has been improved and strengthened throughout time. All of our research techniques are up to par and producing the expected results. This study offers an analysis of the Indian paper currency and suggests a model that may identify the currencies with more accuracy. With the help of CNN model technique's Deep Learning approach, currency identification analysis has become more accurate, faster, and more efficient. What's more, the method is fully automated, requires no human participation, and is minimally difficult. This article outlines a two-pronged approach: first, using Keras to train a DL model; second, using Heroku to host a Flask-based web app. For the purpose of distinguishing between various denominations, our suggested algorithm design and experimentally based findings are helpful for those who are visually impaired.
Index Terms—Currency Prediction, Currency Analysis, Deep Learning, Keras, Convolution Neural Network (CNN), Heroku